0542-4455

Computational intelligence

Also listed as: בינה חישובית

Neural networks, fuzzy logic and evolutionary computation, aimed at physical systems rather than at web data. The strongest bridge in the degree from mechanical engineering into a software or algorithm role.

Semester: Semester B
Weekly hours: 4h
Counts as: Core: systems
Interest areas: Mechatronics and robotics

Comes after

Probability and statistics
Introduction to control

Opens up

0542-4624Lab in robotics and systems control
Graduate machine-learning-for-engineering courses

What it covers

Neural networks

Perceptron and multilayer networks, activation functions
Backpropagation, training and validation, overfitting and regularisation
Radial basis function networks
Introduction to convolutional networks and deep learning
System identification and control applications

Fuzzy systems

Fuzzy sets and membership functions, fuzzification
Rule bases and inference, Mamdani and Sugeno
Defuzzification by centroid and its alternatives
Fuzzy controller design, neuro-fuzzy systems and ANFIS

Evolutionary computation and applications

Genetic algorithms: encoding, selection, crossover, mutation, elitism
Genetic programming, particle swarm and other metaheuristics
Engineering optimisation with metaheuristics
Fault detection, pattern recognition, predictive maintenance, controller tuning

Results worth carrying out

Artificial neuron
$y = \varphi\!\left(\sum_{i=1}^{n} w_i x_i + b\right)$
Gradient descent step
$\Delta w = -\eta \, \frac{\partial E}{\partial w}$
Centroid defuzzification
$z^* = \frac{\int z\,\mu(z)\,dz}{\int \mu(z)\,dz}$

Figures worth knowing

Network architecture diagrams
Activation-function plots
Membership functions and fuzzy control surfaces
Convergence curves for genetic algorithms
Loss against epoch

If you continue to research

Machine learning for physical systems, data-driven control, surrogate modelling, predictive maintenance, physics-informed networks.

If you go to industry

Algorithm engineer, data-driven controls engineer, predictive maintenance and autonomy. Take it with 0571-1818 if you are aiming at software.

Where the subject is practised in Israel

Mobileye
Innoviz
Cybellum
Augury
NVIDIA Israel

Examples of Israeli companies working in this area. Not partners, not sponsors, and not a placement claim.

Jobs this course feeds

Software

MATLAB
Python (scikit-learn, PyTorch, TensorFlow)

Standard texts

Haykin, Neural Networks and Learning Machines
Ross, Fuzzy Logic with Engineering Applications
Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning

Related courses

We also write the summaries.

This map tells you what a course contains. The summaries, the tutoring and the people who already took it are the part you get by joining.

Join CARNOT All courses

The topic outlines are not official syllabi. The university publishes a catalogue blurb per course, not lecture-by-lecture material, so each outline describes what a course of that name, at those hours, with those prerequisites teaches at engineering schools generally, cross-checked against the standard textbook for the subject. Treat it as an informed map, not as a transcript of the lectures.